LearneA

Design and Functionalities of a Hybrid Personalized Learning Electronic Assistant

Nikolaos Mallios and Michael Vassilakopoulos

Dept. of Electrical & Computer Engineering, University of Thessaly, Volos, Greece

Keywords: Personalization, Personalized Learning Assistant, Architectural Model, Design.

Abstract: Throughout the last decade a number of Personalized Learning Environments have been proposed

encompassing various personalization techniques, in order to provide learning material adapted to the

specified user’s requirements. In this study, we illustrate the design and enumerate the basic functionalities

of LearneA, a Hybrid Personalized Learning Electronic Assistant. LearneA practically revisits the way other

existing personalized environments implement the various proposed personalization mechanisms, but

furthermore incorporates a number of innovative characteristics, which have not been implemented yet in

any other learning environment. Our aim is a smart Hybrid Personalized Learning Assistant which will

eventually offer a completely different learning experience. Additionally, a review of previously proposed

Learning Environments is presented with respect to the personalization techniques which they embrace in

order to provide personalized learning content for the user.

1 INTRODUCTION

Although the World Wide Web encompasses of an

extensive number of learning systems, it is however

extremely difficult for someone to isolate and select

the suitable learning material which fits his needs.

Most learning systems offer raw educational

material categorized in chapters which are divided

into sections, and present the specified material in a

modest static way without integrating any kind of

adaptation to the specified user’s requirements.

A number of e-learning systems have been

proposed throughout the years, with many of them

widely adopted and successfully used. Later on,

several researchers investigated the prospective of

proposing a personalized e-learning system which

adapts the learning material offered, to the learner’s

specified needs and characteristics (Dolog, 2004;

Chen, 2005; Huang, 2007). A personalized learning

assistant should provide a flexible and functioning

learning environment which combines specified

personalization techniques and a number of

supporting services.

Traditional e-learning systems were widely

embraced by the academic community throughout

the last decade, due to the necessity of providing an

online heterogeneous learning environment for

students with different learning styles and

capabilities (Dung, 2012). The implementation of a

number of e-learning environments circumvented

many limitations of the classical teaching-in-

classroom (Garrison, 2011; Hsieh, 2011), approach

(formal teaching, tutors’ teaching style, students’

learning capabilities and absorbency of the material

delivered) thus presenting a learning way without the

restrictions of the typical class. Moreover, the

applicability of personalization techniques to the

classic e-learning systems led to personalized learning

environments which include personalized techniques

that fit to the learners’ preferences and needs.

This report records all the necessary

requirements and specifications of LearneA, our

proposed hybrid personalized Learning electronic

Assistant and presents a design architecture for this

assistant. Moreover, all the necessary components

which comprise the overall mechanism are

thoroughly explained and all the personalization

techniques are signalized for both their mechanism

and usage within the assistant.

LearneA practically revisits the way other

existing personalized systems propose and

implement the various personalization mechanisms

which are found in the literature and industry. The

hybrid model of the assistant which we propose

comprises of a number of dynamically evolving

characteristics and components which combine a

number of personalization techniques endorsed in

Mallios, N. and Vassilakopoulos, M.

LearneA - Design and Functionalities of a Hybrid Personalized Learning Electronic Assistant.

In Proceedings of the 18th International Conference on Enterprise Information Systems (ICEIS 2016) - Volume 2, pages 339-344

ISBN: 978-989-758-187-8

339

literature (e.g. ontology-driven profile

personalization, context-aware personalization, query

personalization and recommendation techniques).

To the best of our knowledge, there isn’t any

Personalized Learning Assistant with the

functionalities which are hereby proposed. Our aim

is a smart hybrid Personalized Assistant which will

eventually offer a completely different learning

experience for anyone who wishes to expand his

knowledge perspectives in various learning areas

combined with personalization mechanisms which

will lead him throughout the entire offered course.

In Section 2 we present and categorize related

work, according to the main personalization technique

utilized. In Section 3, we present the architecture and

in Section 4 the functionality of our proposal, while in

Section 5 we present conclusions arising from this

work and we briefly discuss future plans.

2 RELATED WORK

Throughout this section we summarize and present a

number of personalization techniques which are

encountered in literature. Emphasis is given on the

distinct characteristics proposed and the

personalized mechanism each one of them promotes.

An attempt to review Personalized Learning

Assistants with the characteristics that our system

encompasses was made, but there exist no

Personalized Assistants directly focused on learning

material which directly combine wholly the

techniques we propose. Therefore a review is

presented here which includes a number of

Personalized Systems (not Learning Assistants) and

the basic functionalities which they provide

regarding personalization mechanisms, namely the

way these systems provide personalized content

after performing their selected techniques.

Subsequently throughout this section background

material is presented, classified into 4 distinct

categories (i.e. personalization techniques) and a

comparison table is given, highlighting the distinct

characteristics each system delivers in contrast to

our proposed Personalized Learning Assistant.

2.1 Personalization and Context-aware

Systems

Recently a number of context-aware Learning

Systems have been proposed with a number of them

encompassing the idea of a recommendation strategy

which adds the notion of adaptivity and seeks to

present personalized results to the learner using their

system.

One of the most notable proposed systems which

comprises of the ideas of context-awareness but also

includes a recommendation module in order to give

personalized courseware recommendation for the

learner is (Wang, 2011). The authors developed a u-

learning environment where the user of the system

can use a mobile device with RFID technology in

order to connect to the Learning Management

System. The u-learner later on transmits back the

contents of the course to the device, which are

enriched with recommended content with the aim of

the recommendation module. The overall design of

the system is satisfactory and according to the t-test

performed, the results obtained showed a significant

time difference.

Another noteworthy Context Aware Ubiquitous

Learning Environment was proposed in 2006 in

(Yang, 2006) which eventually illustrates and

supports P2P collaborative learning. The use of

ontologies with the aid of Protégé for learner

ontology and service ontology is highlighted in the

construction of the profiles. Both learner and service

ontologies contain surrounding context-awareness

parameters i.e. QoS, environment profile and device

capabilities all part of the so called context

acquisition. Later on context detection and extraction

support the P2P learning environment presented. The

authors demonstrate the use of their proposed system

with a carefully designated scenario.

Furthermore, another two context-aware adaptive

learning systems were proposed in (Chen, 2012) and

(Yaghmaie, 2011). In the first one, the ubiquitous

concept is mainly demonstrated again with the aid of

RFID tags, whereas the overall design architecture

primary consists of 3 distinct modules, the U-

Learning Module, the Teaching Materials

Management Module and the Examination and

Evaluation Module. A series of experiments were

conducted in classrooms and in the Atayal u-

Museum in Taiwan. In the latter one, the proposed

context-aware system is based upon a well-known

open source LMS with the aid of ontologies and

Agents, where 4 types of Agents reside in the overall

system architecture (i.e. Context Management,

Content Selector, Content Organizer and Content

Presenter Agent).

2.2 Personalization and Ontologies for

Describing Profiles or Courses

The use of ontologies is a widely recognized

technique for a number of web applications as well

ICEIS 2016 - 18th International Conference on Enterprise Information Systems

340

as other systems. E-learning systems which are built

upon ontological profiles or courses on a learning

system which are described semantically with the

aid of ontologies are frequently encountered.

One of the many examples of applying

intelligent techniques and semantic web

technologies in e-learning environments is the work

(Gladun, 2009). The main idea is the ontologies

which are built by the tutor (reference ontology) and

by the student (discipline-related ontology) which is

later on compared to the referenced one.

Furthermore, a Semantic Web prototype entitled

M(e)L, is presented which is supported by a number

of Agents -Tutor, Student(s) and Informational-

which interact with each other. Essentially, the Tutor

Agent and Student Agent are not communicating

directly but rather via a broker Agent, the

Informational one, giving promising advantages to

the whole learning experience.

Another noteworthy example of the use of

ontologies in a personalized web search environment

(not a learning one) is described in (Sieg, 2007).

Even though this approach does not directly embrace

the notion of an e-learning environment, the use of

ontological user profiles in order to describe user

context as well as the re-ranking of the results

obtained based on the interest scores in the user

profile, characterizes this attempt for personalized

web search as a noticeably promising one. The same

approach with slight differences is described in

(Mohammed, 2010) where the ontological profiles

are constructed by semantic analysis of the log files.

Finally, another significant methodology which

combines the use of ontologies and recommender

system is the architectural model proposed in

(Shishehchi, 2010). The recommender system

described, consists of two subsystems and their

underlying modules. The use of ontological and

OWL rules demonstrates the rule filtering

recommendation technique.

2.3 Query Personalization

Another interesting approach for providing

personalized results in a learning system is query re-

writing with the aid of a specified user profile,

which is updated dynamically including the user

context and interaction of the system. Such approach

was described for a Learning Management System

in (Paneva, 2006) and in (Koutrika, 2004) for

Database Systems.

For the first one, the authors define a specific

sequence of Learning Objects (LO) and activities

which are tailored to the tutor (rather than the

learner) and moreover introduce the notion of query

personalization by filtering and ranking the results

which are returned by a specific query using the pre-

defined user profile.

In the latter one, the authors present the

conception of query re-writing (query

personalization) by transforming the original query

applied by the user internally (with the aid of their

personalized engine) into another query. For this

purpose they use preferences which are supplied by

the user at an earlier stage and are stored into the

user’s profile. Later on, the extraction of the set P

of top-K preferences take place, derived from the

user profile. These preferences along with the initial

user query are used in a preference selection

algorithm, formulating another personalized query,

thus obtaining different results adaptive to each user.

2.4 Personalization and

Recommendations in Learning

Systems

Finally, a number of proposed architectural models

and the implementation of them embrace the notion

of recommendation. One of the models proposed

upon the SCORM Learning Management System is

the recommendation model LORM (Personalized

Learning Object Recommendation Model) described

in (Wang, 2007). The use of ontologies is present in

the model, in order to identify the Learning Objects

(LO) for the course tailored to a specific learner’s

needs and build a Learning Repository. Later on, a

personal preference pattern is built for each learner

which consists one part of the recommendation

engine, where the second part is the

recommendations based on neighbours’ suggestions.

A thorough and comprehensive survey of

personalized recommender systems, including a

number of learning personalized systems is given in

(Adomavicius, 2005), where in (Khrib, 2008) an

automatic personalization approach is presented,

with the aim to provide recommendations for

learners. The model consists of two modules, an off-

line builder for the models of learner and content,

and an on-line module, which is used to recognize

the student’s needs and apply the various

recommendation techniques.

All the aforementioned papers embrace a number

of personalization techniques (i.e. context-

awareness, query personalization, ontological user

profiles and recommendation algorithms) for

Personalized Learning Systems (Figure 1).

LearneA - Design and Functionalities of a Hybrid Personalized Learning Electronic Assistant

341

Figure 1: Comparison table for personalization features and the aforementioned papers.

3 ARCHITECTURAL MODEL

A considerable number of Personalized Learning

Environments have been proposed and implemented

during the last decade, with each one of them

applying a number of personalization techniques (as

already summarized in Figure 1). Our proposed

architectural model revisits the beforehand

mentioned environments in a significantly different

way, adding functionalities which are not yet

encountered in a Personalized Learning Assistant.

Our vision is to implement an Assistant similar

to requirements and functionalities of the well-

known assistants from Apple Inc. Siri. (Aron 2011)

and Microsoft Cortana (Warren T, 2014). However,

the functionalities of our Assistant will focus on

learning for particular courses.

Indicatively, the three basic characteristics

which differentiate the architectural model proposed

are as follows:

 The Personalized Learning Assistant we envision

uses a text/speech recognition engine (initially

for the English language), where simple/basic

voice commands are recognized and interpreted

by the Assistant. The built-in mechanism of the

speech recognition software “translates” these

commands into basic functionalities for LearneA.

In that way, the learning experience resembles

the use of Digital Assistants like Siri and

Cortana, adding supplementary functionalities to

the Learning Assistant.

 The learning flow is uninterrupted, accessible by

any device the user may possess. The user is able

to continue from the point he stopped with the

aid of the Synchronization Component. This

functionality would be supported by a

web/cloud-based service which is synchronized

by the device which is currently in use and

concurrently synchronizes all the devices

connected to the service. With the support of this

service, the user would be able to continue his

learning experience with all the material being

adapted to the limitations and capabilities of his

device.

 The Personalized Learning Assistant connects to

the user’s life retrieving information and data

from various daily activities. LearneA’s

Behavioural Component will connect to any

social network managed by the user, to his

calendar and mail, thus retrieving all tasks,

appointments and social interaction the user

performs. In that manner, the Assistant is

interrelated with the user’s personal life making

recommendations for the learning progress in an

energetic way (use of automated scripts).

4 FUNCTIONALITY OF THE

COMPONENTS

The overall proposed architecture which is depicted

in Figure 2, incorporates all the necessary basic

components of the design. The main differences

from any other Personalized Learning Environment

have been stated in Section 3 and are namely the

text/speech recognition engine, the Synchronization

Component and the Behavioural Component. All

these characteristics have not yet been implemented

to any other Personalized Learning Assistant, thus

making our proposed model unique and innovative.

A significant role in the overall design of LearneA is

played by the ontological user profile component.

The ontological profile is initially constructed

automatically with basic features of each individual

user (i.e name, age, gender etc.) but is later on

dynamically updated with any useful information the

Assistant captures. Any interaction with the

Learning Assistant is analysed and logged in order

to be integrated into the profile. The use of

ontologies for the modelling and construction of the

user profile is preferred than any other profile

model, since it has been previously selected in many

other research studies and implementation of

Personalized Learning Environments.

Furthermore, the key part of the whole

architecture which plays an essential role in the

personalization mechanism is the Personalization

Component which comprises of 3 distinct

Features/Authors

Wang

2011

Yang

200

Chen C.C.

201

Yaghmaie

2011

Gladun

200

Sieg

200

Mohammed

201

Shishehchi

201

Paneva

200

Koutrika

200

Wang

200

Khrib

200

Context-awareness xx x x x x

Query Personalization xx

Ontologies and/or User Profile xxxxxxxxxx

Recommendation algorithms x xx

Ubiquitous capabilities xx x

ICEIS 2016 - 18th International Conference on Enterprise Information Systems

342

Figure 2: Proposed Architectural Design of LearneA.

integrated sub-components that interrelate and co-

operate in order to provide personalized material,

adapted for each user’s requirements. These are the

context-aware sub-component, the query

rewriting sub-component and the recom-

mendation sub-component respectively.

In the Personalization component a series of

services are performed to provide an expressive

personalization experience. The integration provided

by this component has been implemented in other

Personalized Learning Environments, but our

ambition is to integrate all these mechanisms into a

novel personalized engine.

Specifically, the context-aware sub-component

encapsulates any necessary information surrounding

the learner’s environment (primarily place and time)

as well as other device associated context (i.e. type

of device used to access the Assistant, surrounding

environment of the devices, frequency of usage the

Assistant).

In a parallel manner, the query rewriting sub-

component captures any query the learner poses and

processes it accordingly. Specifically, a query

processing mechanism analyses any given query,

assigning weights to selected keywords, which are

given as input to a modelling query rewriting

algorithm. This algorithm essentially co-operates

consecutively with the recommendation sub-

component. The responsibility of this sub-

component is the selection of appropriate learning

resources to be recommended to the user.

Finally, a fundamental component to the whole

learning model of LearneA’s architecture that we

envision to hold a key role is the Rollback

Mechanism. This mechanism practically monitors

the learner’s activities in conjunction with the

Behavioural Component. During the course of the

whole learning path of the user and in a carefully

automated selected time, the Rollback mechanism

poses a selection of revising questions to the user,

thus determining the overall progress upon the

material. If necessary the mechanism suggests a

revision upon the entire material already covered or

selected parts of it.

5 CONCLUSIONS

In this paper, we presented LearneA, a novel Hybrid

Personalized Learning Electronic Assistant that

encompasses several personalization techniques. The

basic architectural design of our proposed assistant

has been illustrated and the basic functionalities

have been outlined. The proposed system forms an

innovative Personalized Assistant that resembles the

usage and functionalities of other well-known

assistants, like Siri and Cortana, focusing, though,

on learning for particular courses. Furthermore, a

number of other pioneering characteristics of the

proposed system have been illustrated i.e. the

integration of a text/speech recognition engine, the

synchronization component and the behavioural

component. We envision the harmonic cooperation

of all these features within the Assistant, thus

providing a diverse and personalized learning

experience.

Our future plans include the elaboration of the

design of LearneA, its stepwise implementation and,

at the end the delivery of a product which will

enhance in a significant way previous Personalized

Learning Environments.

Personalization Component

Synchronization

Component – Cloud Service

Ontological

User Profile

Behavioral

Com

onent

Rollback

Mechanism

Recommendation

sub-com

onent

Text/Speech

Recognition

Engine

Context-aware

sub-com

onent

Query rewriting

sub-com

onent

LearneA - Design and Functionalities of a Hybrid Personalized Learning Electronic Assistant

343

REFERENCES

Adomavicius, G., & Tuzhilin, A. (2005). Toward the next

generation of recommender systems: A survey of the

state-of-the-art and possible extensions. Knowledge

and Data Engineering, IEEE Transactions on, 17(6),

734-749.

Aron, J. (2011). How innovative is Apple's new voice

assistant, Siri?. New Scientist, 212(2836), 24.

Chen, C. C., & Huang, T. C. (2012). Learning in a u-

Museum: Developing a context-aware ubiquitous

learning environment. Computers & Education, 59(3),

873-883.

Chen, C. M., Lee, H. M., & Chen, Y. H. (2005).

Personalized e-learning system using item response

theory. Computers & Education, 44(3), 237-255.

Dolog, P., Henze, N., Nejdl, W., & Sintek, M. (2004,

May). Personalization in distributed e-learning

environments. In Proceedings of the 13th int. World

Wide Web Conf. on Alternate track papers & posters

(pp. 170-179). ACM.

Dung, P. Q., & Florea, A. M. (2012, May). An approach

for detecting learning styles in learning management

systems based on learners’ behaviours. In Int. Conf. on

Education and Management Innovation (pp. 171-177).

Garrison, D. R. (2011). E-learning in the 21st century: A

framework for research and practice. Taylor &

Francis.

Gladun, A., Rogushina, J., Garcı, F., Martínez-Béjar, R.,

& Fernández-Breis, J. T. (2009). An application of

intelligent techniques and semantic web technologies

in e-learning environments. Expert Systems with

Applications, 36(2), 1922-1931.

Hsieh, S. W., Jang, Y. R., Hwang, G. J., & Chen, N. S.

(2011). Effects of teaching and learning styles on

students’ reflection levels for ubiquitous learning.

Computers & Education, 57(1), 1194-1201.

Huang, M. J., Huang, H. S., & Chen, M. Y. (2007).

Constructing a personalized e-learning system based

on genetic algorithm and case-based reasoning

approach. Expert Systems with Applications, 33(3),

551-564.

Koutrika, G., & Ioannidis, Y. (2004, March).

Personalization of queries in database systems. In Data

Engineering, 2004. Proceedings. 20th Int. Conf. on

(pp. 597-608). IEEE.

Khrib, M. K., Jemn, M., & Nasraoui, O. (2008, July).

Automatic recommendations for e-learning

personalization based on web usage mining techniques

and information retrieval. In Advanced Learning

Technologies, 2008. ICALT'08. 8th IEEE Int. Conf.

on (pp. 241-245). IEEE.

Mohammed, N. U., Duong, T. H., & Jo, G. S. (2010).

Contextual information search based on ontological

user profile. In Computational collective intelligence.

Technologies and applications (pp. 490-500). Springer

Berlin Heidelberg.

Paneva, D., Monova-Zheleva, M., & Zhelev, Y. (2006,

April). Approaches and solutions for personalization

in elearning systems. In proc. of the Third HUBUSKA

<http://www. hubuska. com/> Open Workshop

“Methods and Tools for Development of

Semanticenabled Systems and Services for Multimedia

Content, Interoperability and Reusability (pp. 27-28).

Shishehchi, S., Banihashem, S. Y., & Zin, N. A. M. (2010,

June). A proposed semantic recommendation system

for e-learning: A rule and ontology based e-learning

recommendation system. In Information Technology

(ITSim), 2010 Int. Symposium in (Vol. 1, pp. 1-5).

IEEE.

Sieg, A., Mobasher, B., & Burke, R. D. (2007). Learning

Ontology-Based User Profiles: A Semantic Approach

to Personalized Web Search. IEEE Intelligent

Informatics Bulletin, 8(1), 7-18.

Wang, T. I., Tsai, K. H., Lee, M. C., & Chiu, T. K. (2007).

Personalized learning objects recommendation based

on the semantic-aware discovery and the learner

preference pattern. Journal of Educ. Technology &

Society, 10(3), 84-105.

Wang, S. L., & Wu, C. Y. (2011). Application of context-

aware and personalized recommendation to implement

an adaptive ubiquitous learning system. Expert

Systems with applications, 38(9), 10831-10838.

Warren, T. (2014). The Story of Cortana, Microsoft's Siri

Killer. Retrieved 5/20/2014, from http://www.

theverge. com/2014/4/2/5570866/cortana-windows-

phone-8-1-digital-assistant.

Yaghmaie, M., & Bahreininejad, A. (2011). A context-

aware adaptive learning system using agents. Expert

Systems with Applications, 38(4), 3280-3286.

Yang, S. J. (2006). Context aware ubiquitous learning

environments for peer-to-peer collaborative learning.

Journal of Educational Technology & Society, 9(1),

188-201.

ICEIS 2016 - 18th International Conference on Enterprise Information Systems

344